U.S. patent application number 09/765357 was filed with the patent office on 2001-08-16 for method and apparatus for video compression and restructuring.
Invention is credited to Fu, Hsin Chia, Liang, Yih Woei, Wang, Chein Hsun.
Application Number | 20010014122 09/765357 |
Document ID | / |
Family ID | 21631365 |
Filed Date | 2001-08-16 |
United States Patent
Application |
20010014122 |
Kind Code |
A1 |
Fu, Hsin Chia ; et
al. |
August 16, 2001 |
Method and apparatus for video compression and restructuring
Abstract
A method and apparatus for compressing and restructuring video
signals are provided to promote the usefulness of a video channel,
resulting in the efficacy of accommodating more video programs in a
single video channel. Several video compressing and restructuring
devices are used to directly compress multiple digital video data
streams which are then integrated by a multiplexer. Each of the
video compressing and restructuring devices comprises a plurality
of video compressors for compressing the video codes of the digital
video data stream, and a multiplexer for integrating the video
codes having been compressed by the video compressor. Each video
compressor comprises a trancoder for properly adjusting
quantization scale for further compressing video signals. Based on
the present invention, video signals are directly compressed with
simple method and apparatus of low cost, leading to the realizing
of an analogy Video-On-Demand system (approximate to a
Video-On-Demand system).
Inventors: |
Fu, Hsin Chia; (Madison,
WI) ; Wang, Chein Hsun; (Hsinchu City, TW) ;
Liang, Yih Woei; (Hsin-Chu, TW) |
Correspondence
Address: |
Rosenberg, Klein & Bilker
3444 Ellicott Center Drive-Suite 105
Ellicott City
MD
21043
US
|
Family ID: |
21631365 |
Appl. No.: |
09/765357 |
Filed: |
January 22, 2001 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
09765357 |
Jan 22, 2001 |
|
|
|
09197444 |
Nov 23, 1998 |
|
|
|
Current U.S.
Class: |
375/240.12 ;
375/240.26; 375/E7.016; 375/E7.198; 375/E7.268 |
Current CPC
Class: |
H04N 21/4347 20130101;
H04N 21/47202 20130101; H04N 21/816 20130101; H04N 19/40 20141101;
H04N 21/6377 20130101; H04N 21/234336 20130101; H04N 21/2353
20130101; H04N 21/658 20130101; H04N 21/2385 20130101; H04N 21/6373
20130101; H04N 21/23655 20130101; H04N 21/2365 20130101 |
Class at
Publication: |
375/240.12 ;
375/240.26 |
International
Class: |
H04N 007/12 |
Foreign Application Data
Date |
Code |
Application Number |
Sep 15, 1998 |
TW |
87115328 |
Claims
What is claimed is:
1. A method for compressing and restructuring video signals to
promote the usefulness of a video channel, comprising the steps of:
(a) processing a plurality of digital video data streams to form
digital video data streams each having a desired bandwidth; and (b)
integrating the digital video data streams having said desired
bandwidth into an integrated digital video data stream having a
bandwidth fitting said video channel.
2. The method according to claim 1 wherein step (a) comprises the
steps of: compressing each of said plurality of digital video data
streams into a digital video data stream having a certain
bandwidth; inputting the digital video data streams having said
certain bandwidth to multiplexers for forming digital video data
streams having said desired bandwidth.
3. The method according to claim 1 wherein in step (b) a
multiplexer is used to integrate the digital video data streams
having said desired bandwidth.
4. The method according to claim 1 wherein said digital video data
stream is the video data having been compressed by MPEG2.
5. The method according to claim 1 wherein the bandwidth of said
video channel is the bandwidth of an analog TV video channel of 6
MHz.
6. A device for compressing and restructuring video signals to
promote the usefulness of a video channel, comprising: a plurality
of video compressors each compressing a digital video data stream
applied thereto into a digital video data stream having a certain
bandwidth. a multiplexer receiving the digital video data streams
having said certain bandwidth to provide a digital video data
stream having a desired bandwidth.
7. The method according to claim 6 wherein said video compressor
comprises a trancoder for compressing the bit rate of the digital
video data stream applied thereto.
8. A video compressor for compressing digital video data stream,
comprising: a trancoder for compressing the bit rate of said
digital video data stream; an input buffer and an output buffer
respectively connected to the input pin and output pin of said
trancoder, for temporarily saving the video code of said digital
video data stream.
9. The video compressor according to claim 8 wherein said trancoder
comprises: a delay buffer receiving from the output pin of said
trancoder said digital video data stream which is then adjusted
therein, and generating an overflow signal according to the
overflow status thereof; a variable length decoder for decoding the
video codes of said digital video data stream to obtain numeral
codes; a dequantizer for restoring quantized data; a quantization
scale predictor nonlinearly computing, according to said overflow
signal and the amount of said digital video data stream to be
outputted immediately, to provide an quantization scale; a
quantizer proceeding quantization to provide an output according to
the quantization scale provided by said quantization scale
predictor and the quantized data restored by said dequantizer; a
variable length encoder for encoding the output provided by said
quantizer.
10. The video compressor according to claim 9 wherein said
quantization scale predictor is made of a neural network, and
inputs said overflow signal, the number of bits of current
Macroblock as well as the Macroblocks following thereafter.
11. The video compressor according to claim 9 wherein said
quantization scale predictor provides a signal to indicate a
normalized value of said quantization scale, said normalized value
is multiplied by a constant for obtaining a quantization scale to
be provided to said trancoder.
12. The video compressor according to claim 10 wherein said
quantization scale predictor provides a signal to indicate a
normalized value of said quantization scale, said normalized value
is multiplied by a constant for obtaining a quantization scale to
be provided to said trancoder.
Description
FIELD OF THE INVENTION
[0001] The present invention generally relates to a method and
apparatus for video compression and reformatting, and particularly
to a method and apparatus for enabling existing video channels to
accommodate the transmission of more video programs.
BACKGROUND OF THE INVENTION
[0002] It is getting more and more important to tackle the problem
that the signal flow capacity and communication quality of a video
transmission systems are always limited by channel bandwidth. This
is because of the bulk of data contained in video signal and the
higher communication quality required by video transmission.
Although the number of video channels allowable for cable TV is
more than 120, only very few remain available given that some are
prohibited or are not suitable for using, resulting in extreme
difficulty for acquiring channels for new video programs and new
medium broadcast companies. Due to the limited number of channels
available for application, the increase of number of the programs
in a single channel shall be a good option for expanding the number
of broadcast programs which, however, is limited by channel
bandwidth, leading to a necessity of digital signal compression
technology.
[0003] Shown in FIG. 1 is a conventional digital video signal
transmission and receiving system in which 8 video programs are
merged by multiplexing technology into a single video channel for
transmission. As shown in FIG. 1, the system comprises a sending
station 100 for sending processed video programs which are then
transmitted through cable 200 to a receiving site to be received by
STB (set-top-box) 300 thereat and delivered therefrom to users.
[0004] In sending station 100, a network management and control
unit 110 is used to manage and control a subscriber management unit
112, a multiplex management unit 114, a conditional access unit
116, a multiplex and sever 118, and a 8:1 multiplexer 120; the
input of 110 is connected to an electronic program guide 122 and a
scheduler/trafficker 124 according to which a tape/archive 126
provides programs to be sent directly or through an encoder 128 to
a multiplex and sever 118 which also receives a live video source
130 processed by a real-time encoder 132. Multiplex and sever 118
in turn sends a video signal to be processed by the 8:1 multiplexer
120 for merging into a single channel and then being processed by a
modulator 140 in order to transmit through cable 200.
[0005] STB 300 receives, through a cable interface 302, from cable
200 a signal which has frequency reduced by a tuner 304 and is then
demodulated by a demodulator 306 into a MPEG2 video signal
consisting of 8 programs, and is demultiplexed afterwards by a 1:8
demultiplexer 308 into individual video signals for being applied
to a bus 310 connecting a Direct Random Access Memory (DRAM) 312
and a flash memory 314 for saving data. Based on the system, the
video programs selected by users are retrieved and saved in a DRAM
316 while the others are ignored. The signals contained in DRAM 316
are decoded by MPEG2 decoder 318, with digital video signals
(digital video data stream) and digital audio signals individually
inputted to a video digital-to-analog converter 320 and an audio
digital-to-analog converter 322 to be respectively converted into
analog video signals and analog audio signals for outputting to an
ordinary TV for displaying. The 1:8 demultiplexer 308 also connects
a infrared ray receiver (IR) 324 which is used by users to select a
desired program through a remote controller.
[0006] Given that the bandwidth of a channel for current TV systems
is about 6 Mhz with transmission speed at about 27 Mbps, and that a
MPEG2 system is adopted, the digital signals will usually be
provided (by most of MPEG2 Encoder, for example) with an average
output speed of 3.3 Mbps. With 3.3 Mbps.times.8=26.4 Mbps<27
Mbps (equation 1), it can be seen at most 8 programs can be
accommodated in a channel, e.g., only 8 programs can be broadcast
simultaneously through one channel even though a MPEG2 system is
used, thereby the number of increased programs is far beyond
significant given that the number of available channels is so
limited.
[0007] FIG. 2 shows a video signal obtained from MPEG2 compression,
most of which are distributed in a small range of bandwidth, with
scarce explosion 402 and swiftly moving rapid pan of high detail
404, implying feasible further compression.
[0008] FIG. 3(A) and FIG. 3(B) illustrate encoding and decoding
algorithm of MPEG2. As can be seen in FIG. 3(A), a MPEG encoder
comprises a discrete cosine transform unit 502, a quantizer 504,
and a variable length encoder 506. Usually a video signal is
converted through the three devices into a bit stream (digital
video data stream) to be sent to user sites through a modulator and
transmission medium. To reduce the bulk of signal flows,, many
frames in MPEG2 system are transmitted on the basis of the
difference between two successive frames, therefore a MPEG2 encoder
further comprises a motion compensation unit 512 and a motion
estimation unit. Due to the need that the two devices must operate
with video signal data, a dequantizer 516 and an inverse discrete
cosine transform unit 518 are further required. The final output is
a MPEG2 bit stream (digital video data stream).
[0009] FIG. 3(B) illustrates the operation algorithm of a decoder,
which reverses the operation shown in FIG. 3(A), i.e., the MPEG2
bit stream (digital video data stream) outputted by the encoder in
FIG. 3(A) is inputted to the decoder in FIG. 3(B), and processed by
a variable length decoder 522, a dequantizer 524, and an inverse
discrete cosine transform unit 526, as well as a motion
compensation unit 528, to eventually obtain a restored video signal
as its output.
[0010] When proceeding quantization, the bulk of video data signal
may be reduced by lowering quantization level. Although lowered
quantization level naturally reduces quantized data signal, it
leads to a drawback that the quality of video frames is
lowered.
[0011] Paik suggested, in U.S. Pat. No. 5,216,503, a multi-channel
video compression system using a statistical multiplexer to
integrate multiple video programs in a conventional video channel.
To avoid the unnecessary waste resulting from too big instant
bandwidth of a single program, a buffer controller is used to
generate, when the total bandwidth of these programs exceeds system
capacity, a signal for requesting the quantizer to adjust
quantization level so that the bandwidth is lowered.
[0012] When the aforementioned patent was filed, digital video
signal standard had not been established, therefore its quantizer
was designed for digitizing video signal (similar to MPEG).
Nowadays some digital video signal standards such as
ISO/IECJTCI/SC29/WG11 for MPEG2 have been established, thereby most
of the video contents are processed according to these standards,
resulting in a necessity of converting digital video contents into
analog contents if the aforementioned patent is to be applied,
leading to the need of extra decoding devices and extremely long
operating time.
[0013] It can be seen now that a practicable method and apparatus
for integrating multiple programs in a conventional video channel
can be adopted only if it fits the existing video system and
maintains the quality of video frames. The requirement, however, is
beyond the capacity of conventional arts.
SUMMARY OF THE INVENTION
[0014] An object of the present invention is to provide a method
and apparatus for integrating multiple video programs in a video
channel.
[0015] When bandwidth is extremely limited, digital video signal is
further compressed according to the present invention under the
condition that it is not to be sensed by the eyes of human being,
leading to more efficient utilization of existing channels. It is
therefore another object of the present invention to provide a
method and apparatus for compressing and restructuring video
signals.
[0016] Another further object of the present invention is to
promote operation efficiency of a video system by enabling a single
channel to accommodate more video programs.
[0017] The other further object of the present invention is to
provide a method and apparatus for directly compressing video
signals to realize a real time video system.
[0018] Furthermore, digital video signals (digital video data
stream) can be directly compressed according to the present
invention to enable a single channel to accommodate more video
programs, therefore it is also an object of the present invention
to provide a video compressor and a method for compressing digital
video data, as well as a trancoder and associated method for
compressing digital video data.
[0019] The trancoder suggested by the present invention is
characterized in that a better quantization scale can be achieved
by determining a new quantization scale when quantizing data. It is
therefore also another object of the present invention to provide a
neural network quantization scale predictor for determining an
optimum quantization scale.
[0020] The compression of digital video signal suggested by the
present invention is characterized in that the quantization level
for the areas of a video frame which are less sensitive to human
eyes is reduced while the quantization level for those which are
sensitive to human eyes is maintained
[0021] In an embodiment of the present invention, multiple digital
video compressing and restructuring devices (or called Q-mux) are
used to directly compress digital video signals (digital video data
stream) which are then integrated by a multiplexer; each digital
video compressing and restructuring device has a multiplexer to
restructure digital codes (digital codes of the multiple digital
video signals) having been compressed by video compressors (or
called Q-presser); each video compressor comprises at least a
trancoder to reduce the quantization level for the areas of a frame
which are less sensitive to human eyes, in order to further
compress digital video signal.
[0022] The present invention may best be understood through the
following description with reference to the accompanying
drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
[0023] FIG. 1 shows a conventional video transmission and receiving
system wherein 8 video programs are merged in a single channel.
[0024] FIG. 2 shows video signal bandwidth distribution of a
conventional MPEG2.
[0025] FIG. 3 shows an encoder and decoder of MPEG2, among which
FIG. 3(A) shows the encoder and FIG. 3(B) shows the decoder.
[0026] FIG. 4 illustrates, on the basis of video bandwidth
distribution, the video compression algorithm suggested by the
present invention.
[0027] FIG. 5 illustrates an embodiment of the present invention
allowing up to 24 video programs to be accommodated in a single
channel.
[0028] FIG. 6 illustrates an embodiment of a trancoder suggested by
the present invention.
[0029] FIG. 7 illustrates an embodiment of a quantization scale
predictor suggested by the present invention, which is achieved by
a neural network of 3 layers.
[0030] FIG. 8 shows an embodiment of a video transmission and
receiving system suggested by the present invention.
[0031] FIG. 9 shows an embodiment of a video-on-demand analogy
system (approximate to a video-on-demand system) suggested by the
present invention.
Reference Numerals
[0032] 100 sending station
[0033] 110 network management and control unit
[0034] 112 subscriber management unit 112
[0035] 114 multiplex management unit
[0036] 116 conditional access unit 116
[0037] 118 multiplex and sever 118
[0038] 120 8:1 multiplexer 120
[0039] 122 electronic program guide 122
[0040] 124 scheduler/trafficker 124
[0041] 126 tape/archive 126
[0042] 128 encoder
[0043] 130 live video source 130
[0044] 132 real-time encoder 132
[0045] 140 modulator
[0046] 200 cable
[0047] 300 STB (SET-TOP-BOX)
[0048] 302 cable interface
[0049] 304 tuner 304
[0050] 306 demodulator
[0051] 308 demultiplexer
[0052] 310 bus
[0053] 312 Direct Random Access Memory (DRAM)
[0054] 314 flash memory
[0055] 316 DRAM
[0056] 318 MPEG2 decoder
[0057] 320 video digital-to-analog converter
[0058] 322 audio digital-to-analog converter
[0059] 324 infrared ray receiver (IR)
[0060] 402 explosion
[0061] 404 rapid pan of high detail
[0062] 502 discrete cosine transform unit
[0063] 504 quantizer
[0064] 506 variable length encoder
[0065] 512 motion compensation unit
[0066] 516 dequantizer
[0067] 518 inverse discrete cosine transform unit
[0068] 522 variable length decoder
[0069] 524 dequantizer
[0070] 526 inverse discrete cosine transform unit
[0071] 528 motion compensation unit
[0072] 601-608 digital video compressing and restructuring devices
(or called Q-mux)
[0073] 611-613 video compressors
[0074] 621 trancoder
[0075] 622 input buffer
[0076] 623 output buffer
[0077] 624 disc drives (computer disc drives)
[0078] 625 high speed network
[0079] 631 multiplexer
[0080] 640 Ethernet network switch (etherswitch)
[0081] 650 8:1 multiplexe
[0082] 700 trancoder
[0083] 702 decoder
[0084] 704 encoder
[0085] 712 delay buffer
[0086] 714 quantization scale predictor
[0087] 716 variable length decoder
[0088] 718 dequantizer
[0089] 720 quantizer
[0090] 722 variable length encoder
[0091] 802 input layer
[0092] 804 concealed layer
[0093] 806 output layer
[0094] 901-908 digital video compressing and restructuring
devices
[0095] 910 multiplexer
[0096] 912 modulator
[0097] 914 frequency multiplier (frequency raiser)
[0098] 916 cable
[0099] 918 set-top-box
[0100] 920 TV set
[0101] 930 digital video compressing and restructuring device
[0102] 931 video tape
[0103] 932 compact disc (CD)
[0104] 933 digital video disc (DVD)
[0105] 934 disc (hard or floppy)
[0106] 935 cable system
[0107] 936 satellite antenna
[0108] 937 satellite
[0109] 938-939 satellite antenna
[0110] 940 head-end
[0111] B1.B2 bit stream (digital video data stream)
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0112] According to the present invention, digital video contents
are directly compressed and multiple video programs are merged into
a single video channel. It can be seen from FIG. 4 that most of
video bandwidth ranges below 1 Mbps therefore further exploitation
of bandwidth can be achieved by further compressing digital video
signals (digital video data stream).
[0113] FIG. 5 shows an embodiment of the present invention, which
comprises 8 digital video compressing and restructuring devices
601-608 each including 3 video compressors, such as 3 video
compressors 611-613 included in 601, each of video compressors
611-613 has a trancoder 621 and buffers connected to its input and
output. For example, video compressor 611 includes trancoder 621
for converting video codes of 3.3 Mbps or higher transmission speed
into video codes of 1.1 Mbps. Trancoder 621 has its input and
output respectively connected to input buffer 622 and output buffer
623 each with a memory capacity of 1 Mb for temporarily saving
video signals. The digital video signals (digital video data
stream) retrieved from disc drives 624 are compressed by trancoder
621 to become video codes of 1.1 Mbps. The video compressor may
also receive digital video signals (digital video data stream) from
another kinds of sources such as that video compressor 613 receives
digital video signals (digital video data stream) from high speed
network 625 and compresses the received digital video signals
(digital video data stream).
[0114] 3 video compressor 611-613 output signals to be integrated
by multiplexer 631 to form signals of 3.3 Mbps. 8 digital video
compressing and restructuring devices 601-608 output signals to be
sent to 8:1 multiplexe 650 through etherswitch 640, and then
integrated step by step to form digital video signals (digital
video data stream) of 27 Mbps to be outputted.
[0115] Each video compressor in the embodiment compresses video
signals into video codes of 1.1 Mbps, each of digital video
compressing and restructuring devices 601-608 has 3 video
compressors and has output of 3.3 Mbps, outputs of 8 digital video
compressing and restructuring devices 601-608 fit right in a
channel of 27 Mbps, thereby a single channel can accommodate up to
(3.times.8=)24 video programs which are 3 times what a conventional
system can provide, allowing cable TV companies to have optimum
arrangement in facing clients and video program providers, in order
to maximize the number of programs while minimize the number of
channels.
[0116] It shall be known by those who are skilled in the art that
the video compressor and the video compressing and restructuring
device suggested by the present invention are not limited by the
aforementioned embodiments. Their configuration or design, as well
as constituent number can be modified to adapt to system
requirements, which are not beyond the scope of the present
invention.
[0117] A preferred embodiment of the hardware for the present
invention is that a digital video compressing and restructuring
device comprises a mother board and 3 pieces of Single Board
Personal Computer (SBPC); the mother board comprises Central
Processing Unit (CPU), Read Only Memory, (ROM), Random Access
Memory (RAM), disc drives, and high speed network interface; each
SBPC comprises CPU. ROM, RAM; 8:1 multiplexe 650 can be made of a
CPU (or a computer).
[0118] A preferred embodiment of the trancoder is shown in FIG.
6(a) and FIG. 6(b). FIG. 6(a) briefly illustrates trancoder 700
comprising decoder 702 for decoding inputted bit stream (digital
video data stream) B1 and encoder 704 for receiving the bit stream
(digital video data stream) decoded by decoder 702 and encoding it
into bit stream (digital video data stream) B2. Detailed
description of trancoder 700 is shown in FIG. 6(b) where delay
buffer 712 adjusts inputted bit stream (digital video data stream)
B1 and generates an overflow signal according to its overflow
status; quantization scale predictor 714 estimates, based on
nonlinear algorithm, optimum quantization scale according to the
current overflow status and the video signal segment to be
outputted immediately; variable length decoder 716 restores the
signal produced by a variable length encoder to numeral codes;
dequantizer 718 restores quantized signal; quantizer 720 proceeds
another quantization according to the outputs of quantization scale
predictor 714 and dequantizer 718; its output is processed by
variable length encoder 722 to provide bit stream (digital video
data stream) B2 as an output.
[0119] The trancoder is characterized in that the parts of video
signal which are to be well sensed by human eyes are less
compressed while those which are to be less sensed by human eyes
are more compressed, in order to achieve maximum compression while
maintain frame quality in the range human eyes can tolerate.
[0120] The compression can be easily done by software in a personal
computer for meeting most requirements for video display quality.
The algorithm for compressing data in the present invention is to
determine a new quantization scale when quantizing data, i.e.,
relatively rough quatitization scale is given to the complicated
parts (the parts with roughness not easily sensed by human eyes) of
a frame, while relatively fine quantization scale is given to the
plain parts (the parts with roughness easily sensed by human eyes)
of a frame.
[0121] In the operation of MPEG2 compression, image processing is
done on the basis of basic unit (Macroblock; MB) which comprises
8.times.8 Pixels. The image signal contained in a MB is processed
by a discrete cosine transformation to become a transformation
coefficient C.sub.ij; quantization is one of several main steps in
the MPEG compression of video signals. If transformation
coefficient C.sub.ij is divided by quantization step size, and then
an operation of making integers is applied, quantization levels
L.sub.ij are obtained below 1 L i , j = int [ C i , j q s i , j ] ,
i , j = 1 , , 8 (equation2)
[0122] where q.sub.s is quantization scale, and is an integer
ranging from 1 to 31 in MPEG2; .omega..sub.ij is a quantization
matrix for applying different weighting of processing to the
transformation of different locations, the weighting is established
through observation by human eyes; practically, however, the higher
frequency the transformation coefficient is associated with, the
less sensitivity it has to human eyes, and the corresponding
locations in the matrix have bigger coefficient (less fine), while
the locations corresponding to a transformation coefficient
associated with lower frequency have smaller matrix value which
leads to finer quantization step size, here a is a quantization
constant, and is assigned to equal 2.sup.4.
[0123] Each video frame having a necessity of bit rate trancoding
shall have its frame type remain unchanged, and have the number of
its total bits and each average quantization scale as well as the
number of corresponding bits recorded. Take I Frame for example,
assume the number of bits of a temporarily recorded frame is
B.sup.Prev bits, the bit rate of inputted video signals is R.sub.1
Mbps, and the bit rate of outputted video signals is R.sub.2 Mbps,
the desired number of bits (T bits) of trancoded output for the
frame is obtained according to the ratio between the bit rates as
follows, 2 T = R 2 R 1 B prev (equation3)
[0124] the number T is the desired number of bits set before the
frame is trancoded, and is theoretically an ideal number of bits of
the trancoded output for the frame. The object of controlling bit
rate is to make the number of bits of the trancoded output for the
frame approximate the desired number of bits.
[0125] After calculating the desired number of bits fir a frame,
the Complexity estimation C.sub.j of each MB of the frame is then
computed, and the desired number of bits (T.sup.mbbits) of each MB
is allocated according to the Complexity estimation C.sub.j of the
MB, as shown below, 3 T j mb = C j C 1 + C 2 + + C m T , 1 j m
(equation4) 4 C j 1 = q j ? B prev , j = 1 , , m (equation5)
[0126] where nm is the number of all MBs in the frame, T is the
desired number of all bits in the frame. Computation of
C.sup.i.sub.j is shown by equation 5 where q.sub.j is the
quantization scale of the j th MB of an inputted frame,
B.sup.prev.sub.j is the number of the bits which are in the
inputted frame and are enclosed by the MB. Because the input to the
trancoder is MPEG2 video signals, the encoded data for inputted
video signals can be known when proceeding trancoding, and higher
efficiency and accuracy can be thus achieved by setting desired
number of bits according to the Complexity estimation C.sub.j of
each MB.
[0127] Whenever the trancoding for a MB is completed during the
process of trancoding, the overflow coefficient of virtual buffer
shall be updated as shown by equation 6 below,
d.sup.i.sub.j=d.sup.i.sub.0+B.sup.mb.sub.j-1-T.sup.mb.sub.j-1
(equation 6)
[0128] where d.sup.i.sub.j is the overflow coefficient of virtual
buffer when trancoding the j th row, B.sup.mb.sub.j-1 is the number
of bits of the output for the (j-1) th row, T.sub.j-1.sup.mb is the
desired number of bits computed by equation 4 for the (j-1) th
row.
[0129] It can be seen from equation 6 that d.sup.i.sub.j is
successively accumulated. In case the number (B.sup.mb) of bits of
the trancoded output for each row before the (j-1) th row exceeds
the computed desired number T.sup.mb, d.sub.j.sup.i will gradually
become bigger until Quantization scale gets so big that the number
of outputted bits starts to be smaller than desired number of bits.
This is the time the overflow coefficient begins to fall off.
[0130] In equation 6, d.sup.i.sub.0 is the initial value of
overflow coefficient for I frame, the initial value in the
beginning is 5 d 0 i = q seed 31 (equation7)
[0131] where .gamma. is the value obtained through dividing bit
rate by the number of frames per second, i.e., 6 = 2 bit_rate
frame_rate (equation8) 7 q seed = q 1 exp [ R 1 - R 2 ]
(equation9)
[0132] where q1 is the quantization scale of the first MB of the
first frame, .beta. is a coefficient related to q1 and is used as
the initial value of the overflow coefficient for next I frame. For
P frame and B frame, the steps before computing overflow
coefficient are the same as those for I frame.
[0133] For each MB, the quantization scale predictor suggested by
the present invention can be used to obtain in advance the
q.sub.i.sup.opt (Optimal Quantization scale) given that the current
overflow coefficient d.sub.i-1 and its desired number
T.sup.mb.sub.i of bits are known. The predication based on
d.sub.i-1 and T.sub.i.sup.mb, is usually not good enough, because
the predication for best q.sub.i.sup.opt based on current d.sub.i-1
and T.sub.1.sup.mb may heavily affect the q.sub.i+1.sup.opt for
next MB, such as the case T.sub.i+1.sup.mb becomes verse large
while d.sub.I is not big enough, resulting in a poor scale to
quantize T.sub.i.sup.mb for q.sub.i+1.sup.opt . Observation of more
T.sub.j.sup.mb(j>1) will be more proper for determining
relatively suitable q.sub.i.sup.opt. It must also be noted that the
relations between q.sub.i.sup.opt and d.sub.i-1, T.sub.i.sup.mb ,
T.sub.i+1.sup.mb, . . . are nonlinear, and therefore the
computation for the predication can be based only on experienced
formula associated with complicated computation and accompanied
with inaccuracy It is therefore an object of the present invention
to provide a neural network workable with learning approach in
order to better define the relations between q.sub.i.sup.opt and
d.sub.i-1, T.sub.i.sup.mb, T.sub.I+1.sup.mb, . . . ,.
[0134] FIG. 7 shows a preferred embodiment of a neural network
which is a 3 layer of Multi-Layer Perceptron (MLP). It comprises an
input layer 802, a concealed layer 804, and an output layer 806.
Try each of various different values for d.sub.i-1,
T.sup.mb.sub.i-1, . . . , to find, by human experimentation, a
q.sub.i.sup.opt for best frame performance, and then train the
neural network according to these values. Due to its Generalization
capability, the neural network can make optimum predication for
various cases. It must be noted that the output value of the neural
network ranges between 0 and 1, thereby the outputted
q.sub.i.sup.opt appears as a normalized value which must be
multiplied by a constant.
[0135] FIG. 8 shows an application example of the cable TV
broadcasting and receiving system suggested by the present
invention. Configured on broadcasting site are 8 digital video
compressing and restructuring devices 901-908 forming a single
channel through multiplexer 910, with video output fed to cable 916
through modulator 912 and frequency multiplier 914, for users to
retrieve video programs from set-top-box 918 on remote site and
display the programs on TV set 920. The operation of set-top-box
918 is the same as the set-top-box 300 shown in FIG. 1.
[0136] The present invention' feature of enabling a single channel
to accommodate many programs contributes significantly to the
establishment of a Video On Demand (VOD) system. FIG. 9 shows an
analogy Video On Demand system (NVOD) provided by the present
invention, in which a digital video compressing and restructuring
device 930 as that shown in FIG. 5 is configured on broadcasting
site, and 24 video programs are merged into a single channel. There
can be various options for the source of the video programs, among
which are video tape 931, Compact Disc (CD) 932, compressed video
signals, digital video disc (DVD) 933, and floppy disc 934
containing compressed image, etc. After being integrated by digital
video compressing and restructuring device 930, and broadcast
through cable system 935 or through satellite antenna 936 as well
as uplink satellite 937, these programs can be directly received by
users through satellite antenna 938, or received by cable TV
service companies through satellite antenna 939 and then fed to
cable system 935 via headend 940. Because 24 programs can be merged
in a channel, if a hot program is broadcast through a sub-channel
every 2.5 minutes, by considering 2.5 minutes.times.24=60 minutes
(equation 10), it can be seen that the broadcasting of a movie
based on a NVOD provided by the present invention can proceed with
original video signals of one copy.
[0137] While the invention is described in terms of what are
presently considered to be the most practical and preferred
embodiments, it must be understood that the invention is not
limited to the disclosed embodiment. On the contrary, it is to
cover various modifications and similar arrangements included
within the spirit and scope of the following claims which are to be
accorded with the broadest interpretation to encompass all
modifications and similar structures based thereon.
* * * * *